Parameter Estimation for Electric Motor Condition

Monitoring

Juggrapong Treetronga, Jyoti K. Sinhaa,*, Fengshu Gub, Andrew Ballb

a
School of Mechanical, Aerospace and Civil Engineering,
The University of Manchester, Seckville Street, P.O Box 88, M60 1QD, UK
b
School of Computing and Engineering
The University of Huddersfield, Queensgate, Huddersfield HD1 3DH, UK

*Corresponding Author, E-mail: Jyoti.Sinha@manchester.ac.uk

ABSTRACT

This paper presents parameter identification technique to quantify the faults in motor
condition monitoring. Genetic Algorithm (GA) has been used as a key technique to
estimate the motor parameters. The zero-sequence voltage equation for the stator has been
used as a model to estimate motor stator parameters – the stator resistance and the stator
leakage inductance. The comparison of the parameter estimation by the earlier Rrecursive
Least Square (RLS) method and the proposed GA technique has been discussed. The GA
technique shows better accuracy in the estimation. The estimation has been tested on
both simulations and a real test motor.

KEYWORDS
Condition Monitoring, Parameter Identification, Induction Motor, Genetic Algorithm,
Stator Fault
1. INTRODUCTION

The induction motors are the most widely used motors among different electric motors
because of their high level of reliability, efficiency and safety. However, these motors are
often exposed to hostile environments during operation which may leads to early
deterioration of motors i.e., development of faults. If the fault is not detected at the early
stage, the problem may become serious such as secondary damages to downstream
equipments, and often leads to the unexpected breakdown. To reduce such negative
impacts from the motor fault, the number of the condition monitoring techniques has been
suggested in the literature for the early fault detection and diagnosis so that the remedial
action can be done in much planned way to reduce the machine downtime and to maintain
the overall plant safety. Heng et al. [1] and Jardine et al. [2] gave the review on the
condition based maintenance (CBM) of the rotating machines. The general and/or
accepted practice to use the vibration diagnosis to identify the faults in the motor [3],
however the limitation of this approach is that it can quantify the extent of the fault in
terms of the motor related parameters. On the other hand, the Motor Current Signature
Analysis (MCSA) is one of the most spread procedures for health monitoring of the motor
since decades. One of the main reasons for using this method is that the other methods
require invasive access to the motor and they also need extra equipment/sensors for
measuring the required signals. The research has been progressed in mainly two
directions using the stator phase current and voltage signals – the detection of faults [4-
16] and the quantification of the faults by the motor parameters estimation [17-27]. First
one is important for the quick health assessment on routine basis, however the later one
useful to know the extent of the faults so that remedial action can be done quickly. Hence
the present study is related to the motor parameter estimation.

It has also been observed that 30-40% of all the recorded faults are generally related to
the stator of the motor [4], hence the estimation of parameters related to the stator has
been considered using the motor phase current in the present study. Several methods have
been suggested to estimate the motor related parameters with their relative advantages
and limitations [17-27]. Recursive Least-Square (RLS) has been applied to estimate
motor parameters [17-19]. Treetrong et al. [17] have also used the RLS method to
estimate the stator related parameters. Horga et al. [18] have used the RLS method to
estimate the squirrel-cage induction motor related parameters. They used algorithm of the
continuous parametric model of the induction motor for this purpose. The model was
based on a technique that used the Poisson moment functional theory. The RLS method
was also applied to determine the rotor resistance, self-inductance of the rotor winding,
and the stator leakage inductance of a three-phase induction machine [19]. Extended
Kalman Filter (EKF) is another optimization technique that has been proposed earlier to
determine the motor parameters [20-21]. Velazquez et al. [20] have used the EKF method
to identify the speed of an induction motor and rotor flux based on the measured
quantities such as stator currents and DC link voltage. In another study [21], the EKF
method has been used to estimate the speed of induction motor from speed-sensorless
field-oriented control and direct-torque control of induction motors.

Genetic Algorithm (GA) is one of intelligent search technique to find optimized solution.
It has also been used to determine the motor parameters. It is because the GA method
efficiently handles both linear and non-linear equations and the estimation shows
generally high accuracy [23-27], compared with the conventional recursive method. In
fact, in absence of the actual values of the rotor and stator related parameters in healthy
condition, one can estimate these parameters using the motor specifications generally
listed in the nameplate by the earlier studies based on the GA method [23-24]. Huang et
al. [25-26] have proposed a GA based method to estimate both the rotor and stator
parameters using a motor model in the Park’d-q reference frame. The estimation method
uses fewer measurements but was just validated on simulation and it requires data during
machine transient operation which restrict the practical use of this method. Abdelhad et
al. [27] have used a model based on the single-phase equivalent of Park’model to
estimate the motor parameters. But, the model requires data from off-line tests to estimate
parameters. Hence, the proposed GA method for the stator parameters estimation is
different from the earlier studies [23-27]. The present study has used a new scheme on the
parameter estimation using 3-phase current and voltage signals and rotor speed during
normal motor operation. It is practically more viable for any condition monitoring method
as there is no requirement of the machine transient operation and the off-line tests. The
zero-sequence motor model derived from the unbalance stator voltage for the induction
motor has been used to identify stator parameters – the stator resistance Rs and leakage
inductance Lls . The method has initially been validated on number of simulations and
then tested on the experimental motor with the stator faults only. The advantage of the
proposed GA method over the RLS method in the stator parameters estimation has also
been brought out.

2. ZERO-SEQUENCE MOTOR MODEL

The purpose is to estimate the stator parameters, hence the stator zero-sequence voltage
model for the induction motor has been used. The model can be presented as

dis 0
v s 0  Rs is 0  Lls (1)
dt
where v s 0 is homopolar voltage of stator
i s 0 is homopolar current of stator
Rs is stator resistance
Lls is stator leakage inductance
The homopolar current and voltage can then be calculated by
1
is 0  (is1  is 2  is 3 ) (2)
3
1
vs 0  (v s1  v s 2  v s 3 ) (3)
3
where i s1 , i s 2 , i s 3 are the terminal current of an induction motor
v s1 v s 2 v s 3 are the terminal voltage of the induction motor

The parameters - stator resistance ( Rs ) and stator leakage inductance ( Lls )- have been
estimated from this model. The equation can be arranged into following form for the GA
method.
 dis 0
i s 0  (v s 0  Lls ) / Rs (4)
dt

where i s 0 is estimated zero-sequence current. The input data of the model is stator
voltage. The GA technique is now applied to search the best parameters by comparing the
estimated and measured zero-sequence current.

3. PARAMETER ESTMATION GA METHOD

A Genetic Algorithm (GA) is a search technique used in computing to find solution in

optimization problems [28]. It applies the principles of evolution found in nature to the
problem of finding an optimal solution. In a "genetic algorithm," the problem is usually
encoded in a series of bit strings that are manipulated by the algorithm. Based on
principle of GA optimisation, a parameter scheme for electrical motors can be conducted
based on the process shown in Figure 1 and are summarized in 5 steps.

Figure 1 GA based scheme for the motor parameter estimation

3.1 The Creation of an Initial Population of Parameters ( P00 )

The initial generation P00 has been generated with randomly selected individuals. Each
individual parameter has been constrained by the following condition
Pmin  Pij  Pmax i  1,2,..., n and j  1,2,..., m

where Pmin and Pmax are the limits of the parameter vector value. n I is maximum number
of generation and m is number of parameters or variables. When computation starts, the
initial parameters (decimal number) translated into binary number for the purpose of
generating new population of the parameters.
3.2 Evaluation Operation
At this step, the new populations (binary numbers) translated back to decimal numbers for
the estimation of the model parameters and the error in the objective function.The
objective function is defined as
T max
Err   (i
t 1
s0 (t )  i s 0 (t )) 2 (5)

where i s 0 is estimated zero-sequence current and i s 0 is measured zero-sequence current.

T max is the maximum time length of the data. The program terminates when Err
becomes equal to or less than the set minimal level or when becomes equal to the
maximum generation number ( Ngen ) set in the computational code developed for the
study.

3.3 Selection Operation

The selection process is the next step in the computation that guides GA towards ever-
better solutions. Nowadays, different algorithms are used to select the best individual
value from the estimated population. They are Roulette wheel selection, Tournament
selection, Remainder selection and Uniform selection. Tournament selection has been
applied here to estimate the parameters because it is efficient and easy to implementation

3.4 Crossover Operation

It attempts to combine elements of existing solutions in order to create a new solution,
with some of the features of each "parent”. For example, the elements of the existing
solutions that have been combined in a "crossover" operation as

Parent 1and 2
Child

0 1 0 1 1 1 0 0 0 1 0 0
0 1 0 1 1 1 0 0 0 1 1 1
0 1 0 1 1 1 0 0 0 0 1 1

The probability of the crossover ( Pc ) was assumed to be 0.85 in this study.

3.5 Mutation Operation

It periodically makes random changes or mutations in one or more members of the
current population, yielding a new candidate solution (which may be better or worse then
the existing population members).

Population 1 1 0 0 1 1 1 0 0 0 1 0 1

Population 2 0 0 0 1 1 1 0 0 0 1 1 1
The probability of Mutation ( Pm ) was 0.001 in this study. The computation will get
terminated if the required minimal error in the objective function achieved or the
maximum number of generation achieved.

4. SIMULATION STUDY

A computational code has been developed in the MatLab software code for the estimation
of the stator parameters based on the theory and GA method discussed in Sections 2 and 3
and then applied to the simulated examples. Simulation tests were conducted for 3
different types of induction motors listed in Table 1. The actual values used in the
simulations for Motor-1 to 3 for the parameters - stator resistance ( Rs ) and stator leakage
inductance ( Lls )- are listed in Table 2. To start the computation, the range ( Pmin and
Pmax ) for both parameters was set at 50% of real values, the maximum generation number
50 and the population size equal to 80. The results of the estimated stator parameters for
the 3 simulations are shown in Table 2. The estimated values for the stator resistance
( Rs ) and stator leakage inductance ( Lls ) are close to the actual values (within the error of
6%) of these parameters. Hence the proposed method seems to be estimating the
parameters accurately.

Table 1 The specifications of 3 Induction Motors in simulations

Motor 1 3-phase 4 HP 230 Voltage 4 poles 50 Hertz

Motor 2 3-phase 1 HP 416 Voltage 4 poles 50 Hertz
Motor 3 3-phase 10 HP 220 Voltage 4 poles 50 Hertz

Table 2 The estimated of stator parameters for Motor 1-3

Methods Actual RLS Error (%) GA Error (%)

Parameters value
Motor 1 Rs 2.2530 2.3200 2.97 2.6930 0.7235
Lls 3.1831e-4 3.1799e-4 0.10 3.1986e-4 0.4854
Motor 2 Rs 3.3500 3.4842 4.00 3.385 1.0448
Lls 0.0069 0.0051 26.07 0.06951 0.1614
Motor 3 Rs 0.0453 0.0344 24.13 0.0480 5.899
Lls 2.4669e-4 6.5979e-4 167.4 2.4335e-4 1.3560
Unit: Rs Ohm (Ω), Lls Henry (H)

The results of the RLS method [17] are also listed in Table 2 for comparison with the
actual values and the GA results. It can be seen that the parameter estimation by the GA
show higher accuracy than the RLS. The estimation process by the proposed GA method
has been further refined by a combination scheme. In this scheme, the values of the
parameters estimated by the RLS method [17] have been used as the initial population for
the present GA method. The results are shown in Table 3. It shows that the results are
more accurate (error within 1%) compared to the earlier GA estimation in Table 2.

Table 3 The estimated stator parameters for Motor 1-3 for the Combined Approach (RLS
results used as the initial populations)
Parameters Actual value GA Error (%)
Motor 1 Rs 2.2530 2.2530 0.00
Lls 3.1831e-004 3.18e-004 0.09
Motor 2 Rs 3.3500 3.3510 0.03
Lls 0.0069 0.0067 0.03
Motor 3 Rs 0.0453 0.0457 0.88
Lls 2.4669e-004 2.4679e-004 0.04
Unit: Rs Ohm (Ω), Lls Henry (H)

5. EXPERIMENTAL STUDY

Having validated the proposed GA method on the simulations, the method has now been
tested on the experimental cases. The experimental setup is shown Figure 2. The setup
consists of an induction motor with load cell with a facility to collect the 3-phase current -
voltage signals and rotor speed decoder data directly to the PC at the user define sampling
frequency. The technical specifications of the induction motor used in this experiment are
listed in Table 4. The stator of the motor can be adjusted to 4 different conditions: open
circuit (healthy condition), 5 turn shot circuit, 10 turn short circuit, 15 turn short circuit,
which simulate different levels of the stator fault. Hence the experiments were conducted
for these 4 different conditions at full load condition. The data were collected at the
sampling frequency of 10240 samples/s.

Figure 2 The experimental setup

 Table 4 The specification of the Induction Motor used in the experiment

Phase Power Voltages Currents Frequency PF RPM

Motor 3 4 kW Δ230/Υ400 12.3/7.1 50 0.75 1400

Table 5 The estimated stator parameters for the experimental Motor at different stator
fault levels

Method RLS Method GA Method

Parameter Rs Lls Rs Lls
Open Circuit 1.5830 0.00623 1.5833 0.00621
(Healthy Condition)
5 Turns 0.8113 0.00641 0.8732 0.00630
10 Turns 0.5581 0.00631 0.5652 0.00640
15 Turns 0.3738 0.00567 0.3436 0.00530
Unit: Rs Ohm (Ω), Lls Henry (H)

The experimental results are shown in Table 5 when the initial populations from the RLS
method [17] were used. The decrease in the stator resistance, Rs , indicates extent of open
fault in the stator from the condition 1 to 4 respectively. Typical stator parameters
estimation process by the proposed GA method with generations for the healthy Motor is
shown in Figures 3-4. Figure 5 shows the minimization process (i.e., Objective function,
Err). It can be seen that no divergence has been occurred once the objective function
finds its minima which indicates the robustness of the proposed algorithm.

Figure 3 Stator resistance estimation with generation for the healthy experimental Motor
 Figure 4 Stator leakage inductance estimation with generation for the healthy
experimental Motor

Figure 5 The Objective function with generation for the healthy experimental Motor

6. CONCLUSIONS

The GA method has been proposed to estimate the motor stator parameters using the
Zero-Sequence Motor Model and the measured speed, 3-phase currents and voltages data.
The parameter estimation by the GA shows better accuracy than RLS. The use of initial
populations from the RLS method further improves the accuracy in the parameters
estimation. The method has been successfully tested on the simulated examples and then
applied to the experimental example of an induction motor with 4 levels of the stator
fault. The experimental study indicates that both the stator parameters (Stator Resistance
and Stator Leakage Inductance) decrease with increase in the stator fault. There are
several common motor faults, e.g., loose electrical connections, short-circuits and
imbalanced supply may also be detected by checking the change in stator resistance in a
similar manner, but this needs further experiments. However the proposed method has
shown a potential for more accurate and reliable condition monitoring of the induction
motor. The extension of this method to the rotor related faults is underway.
7. REFFERENCES

[1] A. Heng, S. Zhanga, A.C.C. Tan, J. Mathew, Rotating Machinery prognostics: State
of the art, challenges and opportunities, Mechanical Systems and Signal Processing
23(3) (2009) 724-739.

[2] A.K.S. Jardine, D. Lin, D. Banjevic, A review on machinery diagnostics and

prognostics implementing condition-based maintenance, Mechanical Systems and
Signal Processing 20(7)(2006) 1483-1510.

[3] M. Timusk, M. Lipsett, C.K. Mechefske, Fault detection using transient machine
signals, Mechanical Systems and Signal Processing 22(7) (2008) 1724-1749.

[4] S. Nandi, H.A. Toliyat, X. Li, Condition Monitoring and Fault Diagnosis of
Electrical Motors—A Review, Energy Conversion, IEEE Transactions 20(4)(2005)
719 – 729.

[5] R.R. Schoen, T.G. Habetler, Effects of time-varying loads on rotor fault detection
in induction machines, Industry Applications, IEEE Transactions 31(4)(1995) 900
– 906.

[6] A. Bellini, F. Filippetti, G. Franceschini, C. Tassoni, G.B. Kliman, Quantitative

evaluation of induction motor broken bars by means of electrical signature analysis,
Industry Applications, IEEE Transactions 37(5) (2001) 1248 – 1255.

[7] B. Ayhan, M.Y. Chow, M.H. Song, Multiple Signature Processing-Based Fault
Detection Schemes for Broken Rotor Bar in Induction Motors, Energy Conversion,
IEEE Transactions 20(2)(2005) 336-343.

[8] H. Henao, H. Razik, G.-A. Capolino, Analytical approach of the stator current
frequency harmonics computation for detection of induction machine rotor faults,
Industry Applications, IEEE Transactions 41(3)(2005) 801 – 807.

[9] G. Didier, E. Ternisien, O. Caspary, H. Razik, A new approach to detect broken

rotor bars in induction machines by current spectrum analysis, Mechanical Systems
and Signal Processing 21 (2007) 1127–1142.

[10] S.H. Kia, H. Henao, G.-A. Capolino, A High-Resolution Frequency Estimation

Method for Three-Phase Induction Machine Fault Detection, Industrial Electronics,
IEEE Transactions 54(4)(2007) 2305 – 2314.

[11] A.J. Marques Cardoso, S.M.A. Cruz, D.S.B. Fonseca, Inter-turn stator winding fault
diagnosis in three-phase induction motors’ by Park's vector approach,
Energy Conversion, IEEE Transaction 14(3)(1999) 595 – 598.

[12] A. Bellini, F. Filippetti, G. Franceschini, C. Tassoni, Closed-loop control impact on

the diagnosis of induction motors faults, Industry Applications, IEEE Transactions
36(5)(2000) 1318 – 1329.
[13] R.M. Tallam, T.G. Habetler, R.G. Harley, Stator winding turn-fault detection for
closed-loop induction motor drives, Industry Applications, IEEE Transactions
39(3)(2003) 720 – 724.

[14] H. Henao, C. Martis, G.-A, Capolino, An equivalent internal circuit of the induction
machine for advanced spectral analysis, Industry Applications, IEEE Transactions
40(3)(2004) 726 – 734.

[15] G. Aroquiadassou, H. Henao, G.-A. Capolino, Experimental Analysis of the dq0

Stator Current Component Spectra of a 42V Fault-Tolerant Six-Phase Induction
Machine Drive with Opened Stator Phases, IEEE International Symposium on
Diagnostics for Electric Machines, Power Electronics and Drives (SDEMPED
2007), 6-8 Sept. 2007, pp 52 – 57.

[16] J. Treetrong, Jyoti K. Sinha, F. Gu, A. Ball, Bi-spectrum of Stator Phase Current for
Fault Detection of Induction Motor, ISA Transaction 48(3)(2009) 378-382.

[17] J. Treetrong, K. Zhang, Y.E. Fan, F. Gu, A. Ball, Monitoring electric motors based
on parameters identification techniques, Proc. of 2nd World Congress on
Engineering Asset Management and Fourth International Conference on condition
monitoring, Harrogate, 2007, UK

[18] V. Horga, A. Onea, M. Ratoi, Parameter Estimation of Induction Motor Based on

Continuous Time Model, Proceedings of the 6th WSEAS International Conference
on Simulation, Modelling and Optimization, Lisbon, Portugal, September 2006,
513-518

[19] Y. Koubaa, Recursive identification of induction motor parameters, Simulation

Modeling Practice and Theory 12 (2004) 363–381

[20] S.C. Velazquez, R.A. Palomares, A.N. Segura, Speed Estimation for an Induction
Motor Using the Extended Kalman Filter, Proceedings of the 14th International
Conference on Electronics, Communications and Computers
(CONIELECOMP’04), 2004

[21] M. Barut, S. Bogosyan, M. Gokasan, Speed-Sensorless Estimation for Induction

Motors Using Extended Kalman Filters, Industrial Electronics, IEEE Transactions
54(1)(2007) 272-280

[22] S. Bachir, S. Tnani, J.-C. Trigeassou, G. Champenois, Diagnosis by parameter

estimation of stator and rotor faults occurring in induction machines, Industrial
Electronics, IEEE Transactions 53(3)(2006) 963-973.

[23] P. Nangsue, P. Pillay, S.E. Conry, Evolutionary Algorithm for Industrial Motor
Parameter Determination, Energy Conversion, IEEE Transaction 14(3)(1999) 447-
453.

[24] M. Mutluer, O. Bilgin, M. Cunkas, Parameter Determination of Induction Machines

by Hybrid Genetic Algorithms, KES 2007/WIRN 2007, Part I, 2007, 116 - 124
[25] K.S. Huang, W. Kent, Q.H. Wu, D.R. Turner, Parameter Identification of an
induction Machine Using Genetic Algorithms, Proceeding of the IEEE
International Symposium on Computer Aid Control System Design, Hawai,1999,
USA.

[26] K.S. Huang, W. Kent, Q.H. Wu, D.R. Turner, Effective Identification of Induction
Motor Parameters Based on Fewer Measurements, Energy Conversion, IEEE
Transactions 17(1)(2002) 55-60.

[27] B. Abdelhadi, A. Benoudjit, N. Nait-Said , Application of Genetic Algorithm with a

Novel Adaptive Scheme for Identification of Induction Machine Parameters, Energy
Conversion Transactions 20(2)(2005) 248-291.

[28] D.E Goldberg, Genetic Algorithm in Search, Optimization, and Machine Learning,
Addison-Wesley, 1989.